README.md
8.9 KB · 235 lines · markdown Raw
1 ---
2 language: en
3 tags:
4 - exbert
5 license: mit
6 datasets:
7 - bookcorpus
8 - wikipedia
9 ---
10
11 # RoBERTa base model
12
13 Pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in
14 [this paper](https://arxiv.org/abs/1907.11692) and first released in
15 [this repository](https://github.com/pytorch/fairseq/tree/master/examples/roberta). This model is case-sensitive: it
16 makes a difference between english and English.
17
18 Disclaimer: The team releasing RoBERTa did not write a model card for this model so this model card has been written by
19 the Hugging Face team.
20
21 ## Model description
22
23 RoBERTa is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means
24 it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of
25 publicly available data) with an automatic process to generate inputs and labels from those texts.
26
27 More precisely, it was pretrained with the Masked language modeling (MLM) objective. Taking a sentence, the model
28 randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict
29 the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one
30 after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to
31 learn a bidirectional representation of the sentence.
32
33 This way, the model learns an inner representation of the English language that can then be used to extract features
34 useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard
35 classifier using the features produced by the BERT model as inputs.
36
37 ## Intended uses & limitations
38
39 You can use the raw model for masked language modeling, but it's mostly intended to be fine-tuned on a downstream task.
40 See the [model hub](https://huggingface.co/models?filter=roberta) to look for fine-tuned versions on a task that
41 interests you.
42
43 Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked)
44 to make decisions, such as sequence classification, token classification or question answering. For tasks such as text
45 generation you should look at a model like GPT2.
46
47 ### How to use
48
49 You can use this model directly with a pipeline for masked language modeling:
50
51 ```python
52 >>> from transformers import pipeline
53 >>> unmasker = pipeline('fill-mask', model='roberta-base')
54 >>> unmasker("Hello I'm a <mask> model.")
55
56 [{'sequence': "<s>Hello I'm a male model.</s>",
57 'score': 0.3306540250778198,
58 'token': 2943,
59 'token_str': 'Ġmale'},
60 {'sequence': "<s>Hello I'm a female model.</s>",
61 'score': 0.04655390977859497,
62 'token': 2182,
63 'token_str': 'Ġfemale'},
64 {'sequence': "<s>Hello I'm a professional model.</s>",
65 'score': 0.04232972860336304,
66 'token': 2038,
67 'token_str': 'Ġprofessional'},
68 {'sequence': "<s>Hello I'm a fashion model.</s>",
69 'score': 0.037216778844594955,
70 'token': 2734,
71 'token_str': 'Ġfashion'},
72 {'sequence': "<s>Hello I'm a Russian model.</s>",
73 'score': 0.03253649175167084,
74 'token': 1083,
75 'token_str': 'ĠRussian'}]
76 ```
77
78 Here is how to use this model to get the features of a given text in PyTorch:
79
80 ```python
81 from transformers import RobertaTokenizer, RobertaModel
82 tokenizer = RobertaTokenizer.from_pretrained('roberta-base')
83 model = RobertaModel.from_pretrained('roberta-base')
84 text = "Replace me by any text you'd like."
85 encoded_input = tokenizer(text, return_tensors='pt')
86 output = model(**encoded_input)
87 ```
88
89 and in TensorFlow:
90
91 ```python
92 from transformers import RobertaTokenizer, TFRobertaModel
93 tokenizer = RobertaTokenizer.from_pretrained('roberta-base')
94 model = TFRobertaModel.from_pretrained('roberta-base')
95 text = "Replace me by any text you'd like."
96 encoded_input = tokenizer(text, return_tensors='tf')
97 output = model(encoded_input)
98 ```
99
100 ### Limitations and bias
101
102 The training data used for this model contains a lot of unfiltered content from the internet, which is far from
103 neutral. Therefore, the model can have biased predictions:
104
105 ```python
106 >>> from transformers import pipeline
107 >>> unmasker = pipeline('fill-mask', model='roberta-base')
108 >>> unmasker("The man worked as a <mask>.")
109
110 [{'sequence': '<s>The man worked as a mechanic.</s>',
111 'score': 0.08702439814805984,
112 'token': 25682,
113 'token_str': 'Ġmechanic'},
114 {'sequence': '<s>The man worked as a waiter.</s>',
115 'score': 0.0819653645157814,
116 'token': 38233,
117 'token_str': 'Ġwaiter'},
118 {'sequence': '<s>The man worked as a butcher.</s>',
119 'score': 0.073323555290699,
120 'token': 32364,
121 'token_str': 'Ġbutcher'},
122 {'sequence': '<s>The man worked as a miner.</s>',
123 'score': 0.046322137117385864,
124 'token': 18678,
125 'token_str': 'Ġminer'},
126 {'sequence': '<s>The man worked as a guard.</s>',
127 'score': 0.040150221437215805,
128 'token': 2510,
129 'token_str': 'Ġguard'}]
130
131 >>> unmasker("The Black woman worked as a <mask>.")
132
133 [{'sequence': '<s>The Black woman worked as a waitress.</s>',
134 'score': 0.22177888453006744,
135 'token': 35698,
136 'token_str': 'Ġwaitress'},
137 {'sequence': '<s>The Black woman worked as a prostitute.</s>',
138 'score': 0.19288744032382965,
139 'token': 36289,
140 'token_str': 'Ġprostitute'},
141 {'sequence': '<s>The Black woman worked as a maid.</s>',
142 'score': 0.06498628109693527,
143 'token': 29754,
144 'token_str': 'Ġmaid'},
145 {'sequence': '<s>The Black woman worked as a secretary.</s>',
146 'score': 0.05375480651855469,
147 'token': 2971,
148 'token_str': 'Ġsecretary'},
149 {'sequence': '<s>The Black woman worked as a nurse.</s>',
150 'score': 0.05245552211999893,
151 'token': 9008,
152 'token_str': 'Ġnurse'}]
153 ```
154
155 This bias will also affect all fine-tuned versions of this model.
156
157 ## Training data
158
159 The RoBERTa model was pretrained on the reunion of five datasets:
160 - [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books;
161 - [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers) ;
162 - [CC-News](https://commoncrawl.org/2016/10/news-dataset-available/), a dataset containing 63 millions English news
163 articles crawled between September 2016 and February 2019.
164 - [OpenWebText](https://github.com/jcpeterson/openwebtext), an opensource recreation of the WebText dataset used to
165 train GPT-2,
166 - [Stories](https://arxiv.org/abs/1806.02847) a dataset containing a subset of CommonCrawl data filtered to match the
167 story-like style of Winograd schemas.
168
169 Together these datasets weigh 160GB of text.
170
171 ## Training procedure
172
173 ### Preprocessing
174
175 The texts are tokenized using a byte version of Byte-Pair Encoding (BPE) and a vocabulary size of 50,000. The inputs of
176 the model take pieces of 512 contiguous tokens that may span over documents. The beginning of a new document is marked
177 with `<s>` and the end of one by `</s>`
178
179 The details of the masking procedure for each sentence are the following:
180 - 15% of the tokens are masked.
181 - In 80% of the cases, the masked tokens are replaced by `<mask>`.
182 - In 10% of the cases, the masked tokens are replaced by a random token (different) from the one they replace.
183 - In the 10% remaining cases, the masked tokens are left as is.
184
185 Contrary to BERT, the masking is done dynamically during pretraining (e.g., it changes at each epoch and is not fixed).
186
187 ### Pretraining
188
189 The model was trained on 1024 V100 GPUs for 500K steps with a batch size of 8K and a sequence length of 512. The
190 optimizer used is Adam with a learning rate of 6e-4, \\(\beta_{1} = 0.9\\), \\(\beta_{2} = 0.98\\) and
191 \\(\epsilon = 1e-6\\), a weight decay of 0.01, learning rate warmup for 24,000 steps and linear decay of the learning
192 rate after.
193
194 ## Evaluation results
195
196 When fine-tuned on downstream tasks, this model achieves the following results:
197
198 Glue test results:
199
200 | Task | MNLI | QQP | QNLI | SST-2 | CoLA | STS-B | MRPC | RTE |
201 |:----:|:----:|:----:|:----:|:-----:|:----:|:-----:|:----:|:----:|
202 | | 87.6 | 91.9 | 92.8 | 94.8 | 63.6 | 91.2 | 90.2 | 78.7 |
203
204
205 ### BibTeX entry and citation info
206
207 ```bibtex
208 @article{DBLP:journals/corr/abs-1907-11692,
209 author = {Yinhan Liu and
210 Myle Ott and
211 Naman Goyal and
212 Jingfei Du and
213 Mandar Joshi and
214 Danqi Chen and
215 Omer Levy and
216 Mike Lewis and
217 Luke Zettlemoyer and
218 Veselin Stoyanov},
219 title = {RoBERTa: {A} Robustly Optimized {BERT} Pretraining Approach},
220 journal = {CoRR},
221 volume = {abs/1907.11692},
222 year = {2019},
223 url = {http://arxiv.org/abs/1907.11692},
224 archivePrefix = {arXiv},
225 eprint = {1907.11692},
226 timestamp = {Thu, 01 Aug 2019 08:59:33 +0200},
227 biburl = {https://dblp.org/rec/journals/corr/abs-1907-11692.bib},
228 bibsource = {dblp computer science bibliography, https://dblp.org}
229 }
230 ```
231
232 <a href="https://huggingface.co/exbert/?model=roberta-base">
233 <img width="300px" src="https://cdn-media.huggingface.co/exbert/button.png">
234 </a>
235